ArchSeed
ARCHITECTURE-FIRST DEVELOPMENT FOR AI-ASSISTED BUILDING
I launched the Archseed beta on Peerlist Launchpad this week.
Archseed is a CLI-first workflow for developers using coding agents.It helps turn a vague project idea into clearer scope, non-goals,constraints, architecture decisions, and compact context beforeimplementation starts.
The launch has already led to some useful conversations around a problemI keep seeing with AI-assisted development: the expensive mistakes areusually not syntax mistakes. They are assumptions about scope, boundaries,and decisions that were never made explicit.
The current beta is focused on new and early-stage projects. I’m lookingfor developers building something real with Cursor, Claude Code, Codex,Gemini CLI, OpenCode, or similar tools who are willing to try the CLIand share direct feedback.
One question I’m especially interested in:
What project decision do you most often wish you had made before asking your coding agent to start implementing?
Archseed: https://www.archseed.co/
Peerlist:
https://peerlist.io/mateo_perez/project/archseed
Shipped an updated Archseed beta after developer feedback.
Instead of generating more docs, it now focuses on the decisions an AI coding agent should not invent: V1 scope, non-goals, assumptions, boundaries, edge cases, and a compact agent-context.md.
CLI-first, local, and IDE-agnostic.
Looking for developers to test it on a real upcoming project:
Hey Indie Hackers 👋
I’m a freelance developer, and I’ve been building projects with AI agents.
The biggest change is that I can now work on projects solo that used to require a team. But I’ve also noticed a problem: when the project is not properly planned, AI agents can waste a lot of tokens fixing errors, clarifying requirements, and changing decisions that should have been made earlier.
That’s why I’m building Archseed.
Archseed is an architecture-first, CLI-first tool that helps developers clarify a project before writing code. It asks concrete questions about scope, requirements, constraints, architecture, and implementation decisions, then generates structured files that can be used by the developer or passed to an AI coding agent.
The goal is not to replace coding agents.
The goal is to help them start with better context.
The current workflow can generate artifacts such as:
Archseed is designed to work with any IDE because it starts from the command line. You can use the generated files with Cursor, Claude Code, Gemini, Codex, or any other coding tool.
I’m currently testing the first beta version and looking for developers who:
- Build SaaS products, websites, or MVPs
- Use AI coding agents
- Work independently or in small teams
- Want to reduce ambiguity before implementation
- Care about architecture and project structure
I would especially like feedback on these questions:
1. Are the questions useful before starting a project?
2. Would the generated files help you work with an AI coding agent?
3. Which part of the workflow feels unnecessary or too complicated?
4. What would you expect Archseed to review after implementation?
5. Would you use a tool like this regularly?
The beta is free while I validate the workflow. I’m looking for honest feedback, not just compliments.
If you’re interested in trying it, leave a comment or send me a message and I’ll share access.
Website: https://www.archseed.co/
Thanks for reading!
-
1Hi Mateopp — I use Codex on a JS/TS app and care about the handoff from an agent plan to code review. I can try Archseed on a small non-sensitive project and give concrete feedback on the question flow and review-checklist output. How can I get beta access?
-
1Hi — that would be great, thank you. A small non-sensitive JS/TS project is exactly the kind of beta test I’m looking for. I’d especially value your feedback on whether the question flow helps surface the right decisions before implementation, and whether the handoff is actually useful when working with Codex. You can get started here: https://www.archseed.co/ Once you’ve tried it, I’d love to know: 1. Which question was most useful or unnecessary? 2. What important decision did Archseed fail to ask about? 3. Was the generated context clear enough to use with Codex? 4. Did the review checklist catch anything you would otherwise have missed? 5. Would you use it again for another project or feature? If anything is unclear or breaks, feel free to reply here or email me at m@mateop.com. I’m actively improving the beta based on real usage. I can also give you complimentary beta access in exchange for the feedback. If that works for you, send me the email you’d use for access and I’ll send you a private code.
-
Comment